Prof. Kaizhu Huang
Title: Robust Adversarial Training Improves Trustworthy Artificial Intelligence
Bio:
Kaizhu Huang works on trustworthy AI, machine learning, and pattern recognition. He is Full Professor and Director of Data Science Research Center, at Duke Kunshan University (DKU). Before joining DKU, he was Full Professor at Xi’an Jiaotong-Liverpool University (XJTLU) and Associate Dean of Research in School of Advanced Technology, XJTLU. Prof. Huang obtained his PhD degree from the Chinese University of Hong Kong (CUHK) in 2004. He worked at Fujitsu Research Centre, CUHK, the University of Bristol, the National Laboratory of Pattern Recognition, Chinese Academy of Sciences from 2004 to 2012. He was the recipient of the 2011 Asia Pacific Neural Network Society Young Researcher Award. He published more than 230 international conference papers including 110+ SCI indexed journal papers. He received the best (runner-up) paper or book award eight times at major AI conferences. He serves as an associated editor/advisory board member in a few international journals and book series (e.g. Pattern Recognition Journal, and Neural Network Journal). He was invited as a keynote/tutorial speaker in more than 40 international conferences and workshops.
Abstract:
Artificial Intelligence algorithms including Deep neural networks (DNN) have achieved great success in many applications. However, recent research investigations show that AI algorithms are vulnerable to small perturbations of input data, making them less trustworthy to be applied in many real scenarios. This talk will address fundamentals and theories why the trustworthy issue may happen, and discuss how robust adversarial training can improve the robustness of AI algorithms. While intuitive visualizations and numerical verifications will be presented, outlooks and challenges will be also investigated. This talk will be mainly based on our recent research in trustworthy AI published at ICML, ICCV, ICDM, ACM Multimedia, AAAI, and ECCV.
